How Accurate Is Your Algorithm?

Most people these days have an active online life. In that life, much of the online content is delivered via algorithm. Is that algorithm any good?

Of course, different websites employ different algorithms. The basic idea is the same in each case, however: if you like this thing, you will also like this other thing.

These programs appear just about everywhere online. For example, when you watch a YouTube video, a YouTube algorithm suggests other videos that you may also want to watch.

Not only are these programs ubiquitous, they are crucial to any good website. In the above example, if the YouTube algorithm does not suggest videos you actually want to watch, then you may leave the platform altogether. So, from YouTube’s perspective, it is vitally important that the algorithm work as intended.

What may not be quite so obvious, however, is how important it is to you. There are so many videos available on YouTube that it is impossible to watch them all. In fact, without some way of singling out videos that have some interest to you, you would almost certainly become overwhelmed.

Of course, it is not just YouTube that employs an algorithm to deliver content. Social media sites do it as well. They may suggest other accounts you may want to follow, or posts you may want to see. Online stores use them to practice a modern-day version of an ages old technique of up-selling. Online advertisements, which use cookies to track your browsing habits, can be tailor made to get you to buy something.

In your experience, are these algorithms any good? Does Netflix suggest TV shows and movies that you actually want to watch? Does Amazon recommend products that actually end up in your shopping cart? Is your Facebook feed made up of posts that interest you?

How accurate is your algorithm?

Related questions: What makes us comment on social media? What do you get out of social media? How does technology influence your emotions? How does media manipulate you?

2 thoughts on “How Accurate Is Your Algorithm?”

  1. The main algorithms that offer suggestions of what else I might like to see on the platforms I frequent are close but imperfect. On the whole, they overestimate my assumed preferences. For instance, Apple Music knows I like folk, Americana, and some rap. But in their weekly “New Music” deliveries of what I may like, while the algorithms provide several options of songs within each of the aforementioned genres, it neglects rock selections I might also like. Estimations like this are the same on Facebook and YouTube.

    Specifically, on Apple Music’s algorithm’s accuracy, it correctly delivers songs using minor keys (which I really like), but I only choose to purchase about one in twenty of the songs (or sometimes albums) suggested. That means I get many songs suggested that are okay but are not spot on. The percentage seems close to what a random selection of folk, Americana, and rap would be.

    In general, except for its Reels app, Facebook could do a better job of selecting products, pages, and topics I’d like to follow. As for Instagram (owned by Facebook), I assume it will do a better job, as my use of hashtags will bring more specificity to my liking heirloom vegetable gardening pictures and text.

    On YouTube, the algorithms have a pretty good grasp of artists from whom I like to watch live performances. But, again, it amplifies specific genres while skipping others. Surprisingly, YouTube gets that I like rock but largely neglects folk, Americana, and rap.

    Lastly, all the platforms’ algorithms, except YouTube, feel geared to current recordings or topics even though there is plenty of content from years ago they could suggest.

  2. Most suggestion algorithms seem about 20-30% accurate for me. Netflix does a bit better, perhaps 40%. I think shopping algorithms are the poorest performers, as they have no awareness that if you just bought a thing, you are unlikely to need another of the same anytime soon.

    Pandora did very well and I was able to expand my artist repertoire through them, but I was on free tier and (on Mobile app) really hated the interrupting ads, so I don’t use them much lately. Then again, on Pandora you seed the algo with your current likes. Sort of like YouTube subscriptions being the seed of their suggestions.

    On Reddit I only browse my own chosen subs so I rarely see suggestions from them.

Leave a Reply

Your email address will not be published. Required fields are marked *